RL Techniques for Generating Educational Content
RL approaches can be used to produce educational content such as videos, quizzes, and exercises. This type of content is known as procedural content generation (PCG ). Research works studied the applicability of RL for PCG to generate various levels of racing games [1] and Sokoban puzzles [2,3]. The Monte Carlo tree search (MCTS ) technique is used to generate new tasks in the visual programming field. These generated tasks can be used in various areas such as assigning new practice tasks, including quizzes or homework, to evaluate students’ knowledge. If a student fails, then the student could be assigned a new task to aid in answering the given task. See also [4].
Example:
An examle for generating educational content using RL is RALF, an adaptive reinforcement learning framework based on Cellular Learning Automata to automatically generate content for students with dyslexia, see [4]
Reference:
[1] Gisslén, L.; Eakins, A.; Gordillo, C.; Bergdahl, J.; Tollmar, K. Adversarial reinforcement learning for procedural content generation. In Proceedings of the 2021 IEEE Conference on Games (CoG ), Copenhagen, Denmark, 17–20 August 2021; IEEE: Piscataway, NJ, USA, 2021; pp. 1–8.
[2] Khalifa, A.; Bontrager, P.; Earle, S.; Togelius, J. Pcgrl: Procedural content generation via reinforcement learning. In Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment, Virtual, 19–23 October 2020; Volume 16, pp. 95–101.
[3] Kartal, B.; Sohre, N.; Guy, S.J. Data driven Sokoban puzzle generation with Monte Carlo tree search. In Proceedings of the Twelfth Artificial Intelligence and Interactive Digital Entertainment Conference, Burlingame, CA, USA, 8–12 October 2016.
[4] Fahad Mon, B.; Wasfi, A.; Hayajneh, M.; Slim, A.; Abu Ali, N. Reinforcement Learning in Education: A Literature Review. Informatics 2023, 10, 74. https://doi.org/10.3390/informatics10030074
[2] Khalifa, A.; Bontrager, P.; Earle, S.; Togelius, J. Pcgrl: Procedural content generation via reinforcement learning. In Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment, Virtual, 19–23 October 2020; Volume 16, pp. 95–101.
[3] Kartal, B.; Sohre, N.; Guy, S.J. Data driven Sokoban puzzle generation with Monte Carlo tree search. In Proceedings of the Twelfth Artificial Intelligence and Interactive Digital Entertainment Conference, Burlingame, CA, USA, 8–12 October 2016.
[4] Fahad Mon, B.; Wasfi, A.; Hayajneh, M.; Slim, A.; Abu Ali, N. Reinforcement Learning in Education: A Literature Review. Informatics 2023, 10, 74. https://doi.org/10.3390/informatics10030074
Author of the tip:
Ivo Nowak
HAW Hamburg